November 24, 2025
Will AI make us smarter or dumber? The insights of Klöpping, Scherder and Online Dialogue
Reflection on Klöpping × Scherder by Simon Buil (Data Analyst at Online Dialogue)
How do you implement and use an analytics platform and make sure it scales? Those questions took center stage at the Women in Analytics meetup in Amsterdam on March 9. With Alouette Edens from Schiphol Airport, Rose van Duren of Eneco, and Melody Barlage from Adyen featured a strong lineup, leading attendees through all phases of rolling out an analytics platform.
Alouette Edens, Freelance Digital Business Analyst at Schiphol Airport, kicked off with an answer to the question of how to ensure that tracking data from apps is reliable. This is of course always relevant, but even more so if you have an app on multiple operating systems and users can run multiple versions. So how do you ensure that everything is measured in a consistent way, and you can retrieve the data in an easy way? How do you make sure you have a rebuttal if a Product Owner doubts what the data shows? For that, Alouette presented three solutions.
The first solution is to debug the apps on staging before they go live. Alouette itself mentioned that this is rather obvious, but it can still be tempting to skip this step. Checking all events and parameters before going live can be done, for example, in the real-time snapshots in Google Analytics 4 (GA4). However, its disadvantage is that data remains visible only for a limited time, and the interface is not very user-friendly. It takes many clicks to see the all the data. Therefore, Alouette recommends iOS/Android Debugger by David Vallejo to. You can easily connect a phone to that, after which you can debug a lot easier. That way you catch missing tracking, typos in event names and other errors before going live.
For tracking errors that do slip through this check, Alouette also has a solution: a dashboard that shows inconsistencies between different app versions. Schiphol built a Looker dashboard containing all the events and parameters that came in to GA4. The premise for the iOS and Android apps was that they should both have the same events and the same parameters in tracking. So if the dashboard for a particular event only has data for iOS or only for Android, then something is going wrong in the tracking. The tracking could be completely missing, or it could be a typo. For example, Alouette showed a situation where the event was called ‘flight.search.add_flight’ in one OS, and ‘flight_search.add_flight’ in the other. If such an inconsistency is discovered in the dashboard, that's a red flag for the development team, and it can be fixed.
Alouette Edens during her presentation at Women in Analytics
The final solution is a dashboard that shows which app versions on which operating systems you can use to answer your specific data question. Inconsistencies fiddled with based on the previous dashboard, combined with app users not all using the latest version of the app, make it difficult to select the right data. That's why Schiphol has a second dashboard that shows you if there are inconsistencies based on an entered time period, and the events, operating systems and app versions you want to include in your analysis. Based on that information, you can exclude certain versions or operating systems, or adjust your query, in order to do an accurate analysis that gives a complete picture of app usage.
With these three solutions, which prevent tracking errors on production as much as possible, facilitate detection of problems on production, and ensure that the right data is selected, Alouette now dares to be confident against its Product Owner to say, “Yes, I'm sure my data is correct.”.
Roos van Duren, Digital Product Manager at Eneco told how Eneco became a lot more data-driven in a year and a half. Eneco has had the ambition to go through such a transformation for years, but problems with data quality, data access, data visualization and the way of working stood in the way.
The lack of uniform practices and the teams‘ use of different conventions created a data quality problem. For example, in one screen, it was common for one section to send a ’click‘ action and another to send a ’select" action. To ensure consistency, Roos and her team undertook three actions. First, they set up a central KPI framework, including documentation and agreements on how to measure each of these KPIs. They also created a central measurement plan, which established conventions on such things as naming. Moreover, they made analytics part of each scrum team's Definition of Done: without tracking, nothing was put live.
By ensuring that all data could be found in one place and employees knew how to find it data warehouse could use, the problem with data access was solved. Previously, data was scattered across a variety of systems, making it difficult to track the entire customer journey. Now online and offline data is in one location. All other data sources are locked down to avoid fragmentation. But just bundling all the data is not enough. Education is just as important. That is why Roos and her team set up a large education program, including ongoing GA4 classes to consolidate knowledge. This way, Eneco ensures that everyone knows what data is available and how to use it.
Next up was the problem of data visualization: how do you ensure that complex, combined data remains insightful? To that end, Roos and her team built central dashboards in PowerBI. They put a lot of emphasis on usability, because of course they also wanted colleagues to actually use the dashboards. The results include a dashboard with customer journey insights, where marketing teams can view all funnels and paths with just a few clicks.
Perhaps the biggest challenge was moving from a way of working with occasional data use to one where data is the core. Before the transformation, departments sometimes used data, for example, to verify a Product Owner's idea. The goal was to make the entire organization data-driven. That started with small steps, such as using data to show the development team their impact. They started discussing A/B testing, encouraging devs to guess what the result was. In this way, data became a natural part of their work, and developers became much more a part of the A/B testing efforts. Ultimately, initiatives like these led to big changes: teams prioritized their backlog now based on data. Hippo's (HIghest Paid Person's Opinion) got less influence, and ideas with the most value to the customer and the organization come at the top of the backlog.
With these solutions in the areas of data quality, data access, data visualization and ways of working, skepticism turned into enthusiasm. UX-ers who first felt that data interfered with their creative process now saw the value. Other teams found in data the way to business cases substantiate and prioritize ideas. In the end, Eneco's ambition to go through a data-driven digital transformation turned out to be just fine. Roos summarized all this in 4 tips: define a cross-team data strategy, spend time training colleagues, make data usage as easy as possible with user-friendly dashboards, and put a lot of time into changing the mindset to make data a part of everyone's daily work.
... Just a quick note: we send out a newsletter every three weeks that includes the latest blogs, team updates and, of course, news about the offerings in our academy. Click here to subscribe.
Melody Barlage, Product Lead Business Intelligence at Adyen, closed the afternoon by answering the question: how do you keep a BI platform scalable for 20x? In other words, how do you make sure your platform still works smoothly even when you have 20x more users or 20x more data in the future? This question came to Adyen out of necessity. In 2022, their BI platform had stability issues, access management was chaos, and employees couldn't find the data they were looking for. To bring order to this chaos, Melody contributed three solutions in the areas of technology, automation and people. However, she also admitted that it was not yet a done deal: with so many dependencies, so many contributors, it always remains to find the right balance between rules and freedom.
The first solution was a technological one. Adyen switched from STS to Trino to connect Hadoop to Looker, which provided more stability. They also scaled up from one to seven nodes, which allowed work to be distributed and made the setup less vulnerable. Scheduled tasks and manual queries run on different nodes so they don't interfere with each other. And if one node is down, the other six take over the functionality.
In addition, they automated the granting of access. New employees were automatically granted access to the appropriate data based on their job description. When new customers were brought in, access was automatically granted to the corresponding account managers. Access could also be scheduled in advance. All this together ensured that everyone had access to the right data at the right time, with a lot less manual handling.
The final solution is empowering more people. Before, everyone came to the 12-person BI team with every data question. Now many more people are getting involved. There was an overview of who owns what data, so that questions and requests reach the right people. And that means more employees than just the BI team. In addition, the BI team now shares work for data migration processes with the Looker developers. The BI team provides a robust foundation with good documentation, and the Looker developers complete the migration.
With the steps above, the great chaos of 2022 is over, but there are continuing challenges. First, there are many interdependencies between the BI team, the 180 Looker developers, and the infrastructure team. A total of 200 to 250 employees are involved in one way or another. That requires a lot of alignment and conviction. In addition, having so many involved means that good guidelines are important, and the right channels must be found to communicate those guidelines. Finally, it always remains to find the right balance between rules and autonomy. On the one hand, you want to prevent a bad query from taking the whole system down, but on the other hand, the developers must also have the freedom to experiment and improve. Despite these challenges, Melody is pleased with how her team was able to turn the disarray of 2022 into a scalable and stable BI platform.
The ninth edition of Women in Analytics ended with drinks to network and let all the knowledge gained sink in. Do you want to be there next time? Then become a member of The Women in Analytics NL group at meetup.com. That way you will be kept informed of upcoming events. Also important to know: not only women are welcome.
Nederlands Nederlands
English